Diabetes treatment deintensification: when well-controlled diabetes becomes over-controlled

Diabetes mellitus and its complications, including chronic kidney disease and cardiovascular disease, pose a substantial public health burden. In spite of numerous effective medications for diabetes, the optimal management of diabetes remains elusive. Three large randomized controlled trials of intensive glycemic control (HbA1c of 6.4–6.9% achieved) compared to control arms (HbA1c of 7.0–8.4%) observed only minor cardiovascular benefits from lower HbA1c targets, and one of those trials found increased all-cause mortality in the intensive treatment arm¹. While microvascular complications are consistently reduced by better glycemic control, intensive glycemic control has also been consistently associated with increased risk of hypoglycemia^{1, 2}. Reflecting the heterogeneity of patients with diabetes and the competing risks and benefits of lower glycemic control targets, the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) guidelines for diabetes treatment recommend individualization of glycemic control goals based on patient preferences, comorbidities, severity of diabetes-related complications, and life expectancy³. In spite of the cautious language adopted by the ADA and EASD, a number of recent studies demonstrate potentially avoidable hypoglycemic events⁴, possible overtreatment in vulnerable populations^5–7, and low rates of diabetes medication deintensification in individuals with low HbA1c or at risk for hypoglycemia^{4, 8}. This growing body of literature has brought attention to the morbidity caused by diabetes treatment and motivated further investigation into the causes, consequences of, and strategies to prevent potentially harmful overtreatment.

In this issue of Circulation: Cardiovascular Quality and Outcomes, McAlister and colleagues⁹ report on rates of diabetes medication deintensification in individuals with recently diagnosed type 2 diabetes mellitus. They used retrospective data from Optum Insight from 2004 to 2010 to examine whether HbA1c and/or health status predicted diabetes treatment deintensification. The authors identified individuals with recently diagnosed diabetes, on active diabetes treatment, with an HbA1c measurement while on treatment, and with comorbidity data that allowed assessment of health status. The study participants were classified into four categories of glycemic control based on HbA1c and three categories based on health status (“relatively healthy”, having “multiple comorbidities”, or “frail”). They then examined whether HbA1c category or health status was associated with discontinuation or dose reduction of at least one diabetes medication.

While rates of treatment deintensification were slightly greater in frail adults compared to healthy adults, diabetes treatment was reduced or discontinued in fewer than 25% of all patients, including only 21% of those with an HbA1c below 6%. Overall, the authors observed a statistically significant, but clinically minor, gradient in medication deintensification across HbA1c groups. Furthermore, they did not find an increase in deintensification after the findings of the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial were reported in 2008 showing an increase in all-cause mortality for those in the intensive treatment arm¹⁰. Taken together, the findings of this study confirm prior work in other populations that diabetes treatment deintensification is uncommon, regardless of health status, comorbidities, or level of glycemic control^{4, 8}.

This study provides several insights into the care of patients with diabetes. First, the authors demonstrate that deintensification of glycemic medications is rare in a younger, more recently diagnosed population of individuals with diabetes. In combination with prior studies in Veterans and in older individuals in the National Health and Nutrition Examination Survey, the findings of this study highlight that potential overtreatment is widespread^{5, 8}. Second, the finding that deintensification rates did not change after the results of ACCORD trial were reported in 2008 points to potential delays in adoption of new evidence. Finally, it’s notable that the authors attempt to examine frailty and health status using administrative data. While the study population in this paper is younger than in other recent studies of overtreatment and hypoglycemia, approximately 12% of the participants met the authors’ definition of frailty. Surprisingly, the authors only identified slightly higher rates (21% vs. 17%) of deintensification even in this population less likely to benefit from intense glycemic control and at higher risk of hypoglycemia.

This study adds to the growing body of work describing diabetes overtreatment and suggests that overtreatment is an unrecognized problem across the spectrum of patients with diabetes. A challenge to reducing overtreatment has been the limited evidence base guiding treatment deintensification^{4–8, 11}. Perhaps most critical to the lack of data is the priority in the hierarchy of evidence granted to randomized trials, which focus on treatment efficacy, with less emphasis on adverse events. Furthermore, studying deimplementation is inherently challenging as it requires provider and patient buy-in to the idea that treatment continuation is possibly more harmful than discontinuation. Patients who are comfortable on their diabetes treatment, are empowered by their achievement of goal glycemic control, have embraced the importance of glycemic control to avoid diabetes complications, and have never experienced symptomatic or severe hypoglycemia might be uncomfortable with being asked to stop taking or reduce the dose of glycemic medications for research purposes.

One of the major underlying causes of overly intensive diabetes treatment is that “over-controlled” diabetes was at one time “well-controlled.” Consider the following scenario. A patient with few chronic medical problems is newly diagnosed with diabetes, and treatment goals are established shortly after diagnosis based on an expected life expectancy of decades. The initial glycemic control goals are achieved, earning the label of “well-controlled” and placing the patient and provider at ease. While the patient may require minor, gradual increases in diabetes medications, other medical problems set in, some related to complications of diabetes and others merely the diseases that are part of the natural history of aging. Even if the provider acknowledges the risks of hypoglycemia, “well-controlled” diabetes struggles for prioritization at a clinical encounter at which the patient’s more “active” problems are granted primacy. This scenario demonstrates the need for guidance and evidence to motivate the transformation of the quiescent problem of “well-controlled” diabetes on a crowded problem list into the active problem of “over-controlled” diabetes for which there are specific corrective actions.

Provider inertia, patient satisfaction with care, and the tendency for healthcare preferences to default to the status quo all favor treatment continuation. This reluctance to change course on the part of providers and patients in the setting of “well-controlled” diabetes is in some ways similar to the concepts of “loss aversion” and default effects in healthcare decision-making¹². Furthermore, the framework espoused by ADA guidelines, namely basing personalized diabetes treatment decisions in the context of numerous other patient factors, is in opposition to disease-specific performance measures that reward achieving specific HbA1c values. Therefore, achieving optimal diabetes management, including medication deintensification when appropriate, likely requires a broader reframing of the goals of diabetes treatment for patients, providers, and health systems that incorporates and reflects the dynamic nature of health.

Deintensification is a form of deimplementation; as such, lessons from implementation science may be instructive. A consistent finding from implementation science is that successful implementations are often multilevel¹³. For the case of diabetes-related deimplementation there are patient-level, policy-level, and provider-level interventions that could encourage appropriate deintensification. On the patient level, raising awareness of the harms of diabetes overtreatment, the potential benefits of treatment continuation, and developing strategies for patients to arrive at informed preferences is critical. In this vein, the American Geriatric Society’s Choosing Wisely campaign recommends questioning intensive diabetes treatment in older adults¹⁴. On the policy level, one strategy to motivate active management of diabetes in the setting of an aging and clinically evolving patient is to modify performance measures to reflect appropriate and safe diabetes treatment rather than merely achievement of specific HbA1c targets. Indeed, the National Committee for Quality Assurance Healthcare Effectiveness Data and Information Set (HEDIS) quality measures for diabetes care have been liberalized to reflect the potential harms of overtreatment and now specify a goal of <8% in all individuals and <7% in select populations¹⁵. Finally, for providers it would be helpful if treatment guidelines for diabetes could draw from an evidence base for optimal diabetes management that examines not only efficacy of treatment but also the efficacy of deintensification, and that defines clear parameters that signify when a patient who was “well-controlled” transitions to being “over-controlled.” In spite of the challenges of studying deimplementation, it is encouraging that prospective studies of deintensification are underway that will hopefully yield high priority evidence to guide safe diabetes management¹⁶.

While the personalized diabetes management advised by the ADA is attractive to providers who generally favor a patient-centered approach, such individualized care is often the least supported by existing evidence. A robust literature now demonstrates that diabetes treatment is associated with measurable harm and that treatment is rarely reduced in spite of accumulating risks of adverse events. Creative approaches are needed to address the challenge of developing an evidence base for personalized diabetes management that acknowledges the heterogeneity of the population of patients with diabetes and the dynamic, time-varying balance of net clinical benefits and harms associated with treatment.